Energy storage system (ESS) is a power storage device to increase power utilization efficiency by storing generated electricity and supplying power, when needed. As the number of ESS explosion accidents has been rapidly increasing over the past two to three years, a failure diagnosis study on ESS fire safety should be preceded for its continuous use. When the ESS is operated at high temperature for a long time, aging is caused owing to deterioration, and hence, stable operation is required. Among the recent 28 fires, 15 of them occurred during charging and resting. Because rapid temperature and humidity changes occur during charging and discharging, it is important to predict the trend of change. This study predicts ESS operation patterns through an artificial intelligence-based prediction algorithm, and also diagnoses ESS failures through a supervised learning-based confusion matrix. When a failure pattern is found in the ESS, a method to reduce damage through air conditioner control is studied.